ISO/IEC JTC1/SC22/WG5 N1977 Minutes of Meeting of ISO/IEC JTC1/SC22/WG5 Hosted by NEN, the Netherlands Member Body for JTC1/SC22 NEN, Vlinderweg 6, 2623 AX Delft, The Netherlands June 24 - 28, 2013 List of Participants: John Reid (JKR Associates, UK) convenor Dan Nagle (NCAR, USA) PL22.3 chair Reinhold Bader (Leibniz Supercomputing Centre, Germany) (Monday to Thursday) Mark Batty (University of Cambridge, UK) (Monday to Wednesday) Daniel Chen (IBM, Canada) Malcolm Cohen (NAG, UK) Steve Lionel (Intel, USA) Bill Long (Cray, USA) Nick Maclaren (University of Cambridge, UK) Toon Moene (Gnu Fortran, Netherlands) David Muxworthy (BSI, UK) Van Snyder (Caltech/JPL, USA) Masayuki Takata (Edogawa University, Japan) WG5 papers are referenced as Nnnnn. They are available from ftp://ftp.nag.co.uk/sc22wg5/ PL22.3 papers are referenced as 13-nnn. They are available from http://www.j3-fortran.org/ 1. Opening of the Meeting The meeting opened at 09:00 on Monday 24 June 2013. 2. Opening business 2.1 Introductory remarks from the Convenor The convenor said that the principal objectives of the meeting were: firstly to develop the draft TS on further coarray facilities; secondly to consider the defect reports in document 13-006A; thirdly to construct an initial list of deficiencies and discrepancies in Fortran 2008 to be addressed in the next revision. 2.2 Welcome from the Host Toon Moene, on behalf of NEN, welcomed participants to Delft. 2.3 Local arrangements The host described local arrangements and said that there would be a dinner for participants on the Wednesday evening. 2.4 Appointments for this meeting The drafting committee would be Reinhold Bader, Daniel Chen, Toon Moene, David Muxworthy (chair), Van Snyder and Masayuki Takata. David Muxworthy would act as secretary and John Reid as librarian. 2.5 Adoption of the agenda (N1964) The preliminary agenda was adopted, with the addition of three tutorials (see below). 2.6 Approval of the minutes of the Markham 2012 Meeting (N1926) The minutes of the 2012 meeting were approved. It was agreed that in future minutes would be circulated for approval by WG5 letter ballot approximately one month after each meeting. 3. Matters arising from the minutes The convenor reported that it was no longer proposed that the CD stage be omitted from standards development; in future the CD stage would be optional, as determined by the Working Group. 4. Status of Markham 2012 Resolutions (N1927) There was nothing to report. 5. Reports 5.1 SC22 Matters (Convener) A report of the 2012 SC22 meeting was in document N1940. The meeting had, as an experiment, allowed remote participation. There was then a short discussion as to whether remote participation would be suitable for WG5 meetings. Opinion generally was not in favour. It was reported that the three-year, or three-edition, limit on corrigenda (v. ISO/IEC Directives Part1, 2.10.4) would not to be enforced. 5.2 National Activity Reports (Heads of Delegations) Canada: The Canadian programming language committee had representatives for Ada, C, C++, Cobol and Fortran. It met at IBM. Germany: A set of requirements for the next revision was being developed but was not yet complete. Tobias Burnus had implemented a large part of the interoperability TS for gfortran. Japan: Translation of Fortran 2008 was proceeding slowly due to lack of resources. It would take at least another two years. Netherlands: The Dutch programming language committee, NC381022, had representatives for C, C++, Cobol and Fortran and met twice a year. UK: Report is in N1976. US: Report is in N1974. 5.3 Report from Primary Development Body (NCITS/J3 Chair) Report is in N1974. 5.4 Reports from other Development Bodies (Editors/Heads) TS on further coarray facilities: Bill Long This was dealt with under item 6 below. 5.5 Liaison Reports: NCITS/PL22.11 (C): Dan Nagle There was nothing of note to report. MPI: Bill Long Developments in the MPI 3 Fortran interface looked encouraging although there was a lack of knowledge of Fortran. UPC: Bill Long (for Damian Rouson) UPC 1.3 draft consisted of the language, a required library and an optional library. Some UPC customers were unhappy with the state of UPC. Some had come to Cray asking for new features. There was now a Cray C++ coarray compiler with the Fortran semantics and analogous syntax reproduced in C++ using a class library. No change to the actual C++ compiler was involved. IFIP/WG2.5: Van Snyder No news was available. OpenMP: Bill Long The OpenMP Board were incorporating the capabilities of OpenACC and had almost completed the work. WG23 (Vulnerabilities): Dan Nagle The current draft of the third edition of the Vulnerabilities TS, WG23/n0453, included a Fortran annex. A couple of mistakes had been found in the annex. WG23 were continuing to add new vulnerabilities. WG21 were not interested in contributing an annex for C++. [PL22.3 plenary and subgroup sessions from 10:05 to 16:30] From this point agenda items 6-9 largely proceeded in parallel. 6. Construct the PDTS on Further Coarray Features. 7. Consider the Fortran defect reports (interpretations) in J3-006. 8. Construct an initial list of deficiencies and discrepancies in Fortran 2008 to be addressed in the next revision. 9. Consider a proposal from USA concerning a system of units of measurement in Fortran. The subgroup developing the further coarray TS reported on progress. The subgroup examining proposals for the revision of Fortran 2008 which were in documents 13-244r1 and N1975 reported on their preliminary conclusions (described in more detail below). For ease of reference proposals from 13-244r1 are written as US-nn and those from N1975 as UK-nn. The meeting adjourned at 17:00. ........................................................................... Tuesday Paper 13-271 containing edits to N1967 was approved unanimously; paper 13-272 was withdrawn after some discussion; paper 13-273 was withdrawn after a straw vote 6-1-5 in favour of removing EVENT_TYPE (and keeping LOCAL_EVENT_TYPE renamed as EVENT_TYPE). For the revision of Fortran 2008, a block vote was proposed on the adoption of items US-01, US-03, US-04, US-05, US-08, US-12, US-14, UK-02, UK-03, UK-05, UK-06 and UK-08. Following objections, items US-01, US-14 and UK-08 were withdrawn from the vote. The remaining set was approved by 11-1-0. After further consideration US-01 (extending the G0 edit descriptor) was approved by 11-0-1, US-14 (updating with respect to IEC 60559:2011) by 12-0-0 and UK-08 (new reduction intrinsic) by 10-0-2. The subgroup had recommended approval of all of UK-10 (new obsolesences and deletions) other than subitem 2c (obsolescing implicit interfaces). There were objections to subitems 1a, 1b, 2f and 3. After further discussion parts were voted on separately: the set 2a, 2b, 2d and 2e was approved by 11-0-1, 1a and 1b (new deletions) by 10-1-1 and 2f (obsolescing the FORALL construct) by 11-0-1. A decision on subitem 3 (unobsolescing the CHARACTER* form) was postponed. [PL22.3 subgroup sessions from 10:30 to 13:40] Mark Batty gave a presentation entitled "Relaxed Memory: The Specification Design Space". He described and contrasted two main models of memory specification; this was followed by discussion. The slides are in document N1985. [PL22.3 subgroup sessions and plenary from 15:30 to 16:45] Subitem 3 of UK-10 was reconsidered. The vote was 5-5-2; this was deemed to be disapproval. The meeting adjourned at 17:00. ........................................................................... Wednesday After some short PL22.3 business, consideration returned to revision of Fortran 2008. Subgroup recommended that at most one of US-13 (Conditional expressions and arguments) and UK-01 (Default KINDs for constants and intrinsics) be adopted. Although many users would find UK-01 useful with modern hardware there was concern that decoupling the concept of 'default kind' from 'single precision' could cause difficulties with separate compilation and with links to libraries, and further that UK-01 was on the margins of what was allowed by Markham resolution M6. Nevertheless a vote on which of the two was preferred gave: US-13 4, UK-01 7, equal preference 1. A vote on whether it should be actively pursued gave: yes 2, wait 7, indifferent 3. To clarify the situation with these two papers there was a vote to confirm that the following would not taken up this year, subject to possible review by WG5 in 2014: US-02, US-06, US-07, US-09, US-10, US-11, US-13, US-15, US-16, US-17, UK-01, UK-07 and subitems 2c and 3 of UK-10. This was approved nem con. UK-09 had addressed the same point as US-14 and was discounted. At this point all items other than UK-04 had been decided upon. Papers 13-265r1 and 13-272r1 which detailed updates to N1967 were both approved by unanimous consent. Papers 13-275 and 13-276 were both withdrawn for revision. Van Snyder gave a presentation "Units of Measure in Fortran" which proposed a new TS which would add the ability to check physical dimensions and measurements in Fortran. The slides are in N1970 and a draft TS is in N1969. It was emphasized that these were the result of many years of work and discussion with engineers at JPL. A vote on the subject was postponed to allow further informal discussion. [PL22.3 subgroup sessions and plenary from 11:45 to 16:15] There was a discussion on the CHANGE TEAM construct in N1967. A proposal that the rules be changed so that CHANGE TEAM involves the separate synchronization of each of the teams it creates was approved 12-0-0. A straw vote on a proposal that images of the parent team could access subteams was 5-2-5. Kees Ampt, a member of various programming language committees since the mid 1970s, and especially active in WG5 from 1978 to 1985, gave a presentation on the history of Fortran standardization, including IRTF. The slides are in N1984. [Secretary's note; some related historical information is also in N025, N030, N055 and N075 and in http://www.fortran.bcs.org/standards/meek_hague_1977_report.pdf] The meeting adjourned at 17:20. ........................................................................... Thursday It was proposed that subitem 1 of UK-04 in N1975 be adopted and that subitems 2 and 3 not be addressed in Fortran 201x, subject to possible WG5 review in 2014. This was agreed 12-0-0. Papers 13-276r1, 13-279, 13-280 and 13-282r1 which detailed edits to N1967, were passed by unanimous consent. There was a discussion on changing the requirements for further coarray features. A vote on whether to add syntax to specify a codimension relative to another team was 6-0-6. A vote on whether to allow general access of coarrays across subteams was 6-2-4. A vote on whether FORM SUBTEAM should allow choice of image number was 5-0-6. A proposal that objects should include minimal portable handling of failed images was approved by 9-1-2. The revisions to the requirements are in 13-292r1. [PL22.3 subgroup sessions and plenary from 10:30 to 16:45] There was a discussion on revision of N1925, the WG5 Strategic Plan (agenda item 10). The plan from 2012 had assumed that technical work on the TS on further coarray facilities would be essentially complete when revision of the full language was still in its preliminary stages, enabling it to be incorporated into the revision. This now seemed unlikely. Thus the question was whether, and by how much, the schedule for the latter should be delayed, or whether the two schedules should be uncoupled. Discussion was adjourned to the following day. The meeting adjourned at 17:00. ........................................................................... Friday Reinhold Bader had asked David Muxworthy to act as his proxy in any WG5 votes on Friday. Papers 13-258r1, 13-263r1, 13-273r2, 13-287r1 and 13-290r1 which contain edits to N1967, were passed by unanimous consent. Papers 13-288, 13-289 and 13-291 were withdrawn for reconsideration. There was a discussion on the revision of requirements for the further coarray TS (N1930). It was proposed to delete the requirement that when a block of code is executed on images executing as a team, it should execute on those images as if the program contained no other images. A country vote was: delete the requirement 3 (CA, NL, US); retain it 0; abstain 3 (DE, JP, GB). There was a discussion on future meetings (agenda item 11.1). It was decided not to change the existing proposals for 2014 and 2015. Dan Nagle offered to investigate possibilities in the Boulder area for 2016. This was gratefully accepted. Two alternative plans for the schedule of the further coarray TS were considered (agenda item 10.1). It was decided that decoupling the TS schedule from the main language schedule could be avoided if the former slipped by no more than four months. The agreed schedule is in N1979. There were votes on whether WG5 should apply for a new project to develop a TS on units of measure, following the presentation on Wednesday. The individual vote was: yes 3 - no 7 - abstain 1. The country vote was: yes 0 - no 3 (CA, JP, GB) - abstain 3 (DE, NL, US). [PL22.3 subgroup sessions from 12:40 to 15:00] Papers 13-264r3, 13-281r1, 13-288r1, 13-289r1 and 13-291r1 which contained edits to N1967 were approved. Paper 13-266r2 would not be pursued. [PL22.3 closing business was processed.] 10. WG5 Business and Strategic Plans 10.1 Goals for 2013-2016 These had been discussed earlier in the day. 11. Closing Business 11.1 Future meetings The 2014, 2015 and 2016 meetings had been dealt with earlier in the day. 11.2 Any other business None was raised. 12. Adoption of Resolutions (N1978) Resolutions D1, D2, D7 and D8 were approved by unanimous acclaim. Resolutions D3 to D6 were approved by unanimous consent. 13. Adjournment The meeting adjourned at 15:50 on Friday June 28, 2013.